Artificial intelligence-generated dramas and videos have become a major trend in the content sector.

The format emerged from the niche for AI-generated comic dramas and has been shaped by rapid advances in video generation foundation models. By the second half of 2025, this type of content was already beginning to appear. At the time, some players argued that the upside for AI-generated dramas could rival that of the traditional film and television industry, making them a bigger opportunity than AI comic dramas. That belief drew them in before many of the technical hurdles had been resolved.

Since January, the release of ByteDance’s Seedance 2.0 and Shengshu Technology’s Vidu Q3 has accelerated the sector’s growth. Problems such as the uncanny valley effect in facial expressions, inconsistent character appearances, and choppy action scenes have begun to ease as the underlying models improve.

Zheng Zhong, head of solutions at Vidu, said video generation foundation models have evolved quickly over the past year, moving from consumer-facing applications such as reviving old photos and offering transformation effects to professional tools that can support the production of comic dramas and live-action-style dramas. Their performance in consistency, aesthetics, and semantic understanding has continued to improve, and the technology “has already begun moving into serious creative work.”

Yet the pace of technological change has made the sector’s direction difficult to read. What does a company need to stay competitive in this field? And how far will today’s boom in AI-generated dramas continue into the future?

In March, Yuewen Group held a masterclass on AI drama generation. It was the company’s second such gathering, following a similar session in November 2025, bringing together participants from business, academia, and the media to discuss the sector’s direction.

Across the discussions, participants broadly agreed that AI’s initial impact has been to reduce costs and improve efficiency.

Meng Xiangyun, founder of Aikan, shared at the event that his team had debated its strategy in late 2024, weighing whether to prioritize content or tools. A single conclusion was ultimately reached: “In the end, comic dramas compete on efficiency.”

In Meng’s view, the core competitiveness of live-action-style dramas lies not only in better visuals, but also in whether content can be produced reliably, flexibly, and at scale. That, in turn, makes tool development and project management foundational infrastructure.

That view aligns with a broader argument in the industry: while AI-generated live-action-style dramas cost more to make than AI-generated comic dramas, they can still be produced at less than one-tenth the cost of traditional short dramas, making it possible to produce a title for under RMB 100,000, or about USD 14,500.

That combination of speed and low cost is reshaping supply and strengthening the case for adoption. In the past, limited budgets and shooting constraints meant creators often avoided large-scale spectacle, science fiction themes, and complex ensemble scenes. With AI, those elements have become major audience draws.

Meng said the team now prioritizes “spectacle” when greenlighting projects, focusing first on scenes that were previously difficult to realize, and then building the narrative around them.

As creation becomes less constrained by physical production limits, competition is shifting. With AI, content has moved back to the center of production.

Huang Haonan, founder of Jiangyou Culture, argued that AI can significantly expand production capacity, but it cannot generate unlimited high-quality content. The supply of ideas that creators can provide has limits. By contrast, platforms’ demand for content continues to grow. That makes IP the key resource that determines the ceiling of supply.

In recent years, China’s online literature industry, along with the traditional film, television, and animation industries, has accumulated a large reservoir of IP. But relatively little of it has been developed in a sustained way. As AI lowers production barriers, the efficiency of adapting and developing those IP assets has risen sharply, increasing their value. IP is no longer just source material for adaptation. It is becoming a long-term asset that can be developed continuously and expanded over time.

Deeper development of a single IP could also reshape competition.

Ding Kuan, director at Jiuziyuan AI, said the core of future content competition will shift from rivalry between individual works to rivalry between IP universes.

In an AI-driven environment, creators can continue producing content around a single worldview, using high-frequency updates to build audience stickiness and create long-term value. One of Ding’s works, a film set in an isolated city, performed strongly across major platforms with little more than organic traffic. It also attracted interest from overseas platforms, which Ding sees as early validation of this model.

On that basis, the sector’s structure is also likely to diverge in new ways. On one hand, AI tools are lowering the barriers to creation, giving small teams the ability to produce independently. Ding estimates that teams of five to 15 people could become the main creative unit in the future. On the other hand, companies with deeper IP reserves and the ability to scale production will still be able to serve larger markets through industrial-scale output. The link between the two, he argued, is the continuous development and operation of IP.

As AI gradually makes production capabilities more widely accessible, the truly scarce resource is no longer whether something can be made, but what to make, and what can continue to be made over time. As the source and container of content, IP is becoming more valuable.

For now, the question is whether AI drama generation is still worth exploring.

By market size, the content sector remains in a phase of rapid expansion. At the same time, technological iteration and expanding supply have brought the window for mass adoption closer. 36Kr found that most estimates clustered around three to six months, with the longest forecast at no more than nine months.

Even so, the sector still faces several major challenges. Zheng noted that models still need a better understanding of the physical world, including details such as the weight of objects and the mechanics of movement, all of which directly affect the realism of the final output. In addition, issues such as copyright ownership and content regulation are likely to become more prominent as the industry develops.

KrASIA features translated and adapted content that was originally published by 36Kr. This article was written by Lan Jie for 36Kr.